OpenClaw Voice Assistant: Voice Wake and Talk Mode Setup [2026]

What Are Voice Wake and Talk Mode?

Voice Wake and Talk Mode are two related features that transform OpenClaw from a text-based AI agent into a voice-controlled assistant. Together, they give you an experience similar to Siri, Alexa, or Google Assistant — but powered by the AI model of your choice (Claude, GPT, Gemini) and connected to all of your OpenClaw integrations.

Voice Wake is the always-listening feature. When enabled, your device continuously monitors for a specific wake word — "Hey Claw" by default. When it hears the wake word, OpenClaw activates and enters Talk Mode. This happens entirely on-device using Apple's Speech framework, so no audio is sent to any external server during the wake word detection phase.

Talk Mode is the conversational interface. Once activated (either by Voice Wake or manually), you speak your request and OpenClaw responds through your speakers. Your speech is transcribed on-device, the text is sent to your AI model provider, and the response is converted to speech using text-to-speech synthesis. Talk Mode continues until you say "stop" or a configurable silence timeout is reached.

The key difference between this and consumer voice assistants is scope. Siri and Alexa are limited to their built-in capabilities and supported integrations. OpenClaw Voice gives you access to everything your agent can do — calendar management, email processing, CRM updates, content creation, data analysis, and any custom skills you have installed. You are talking to a fully capable AI agent, not a constrained voice assistant.

Supported Platforms

Voice Wake requires Apple's Speech framework, which is only available on macOS and iOS. On the iOS app, Voice Wake works when the app is in the foreground but not in the background due to iOS power management restrictions.

Talk Mode's push-to-activate variant is available everywhere through the web dashboard. Click the microphone button, speak your request, and the agent responds. This works on any modern browser including Chrome, Firefox, Safari, and Edge.

Setting Up Voice Wake on macOS

Prerequisites: The OpenClaw Mac app (not the npm or Docker installation). macOS 14 Sonoma or later. A working microphone (built-in or external).

Step 1: Grant microphone access. Open System Settings, go to Privacy and Security, then Microphone. Ensure OpenClaw is listed and toggled on. If it is not listed, the first time you enable Voice Wake, macOS will prompt you for permission.

Step 2: Enable Voice Wake. Click the OpenClaw menu bar icon, then open Preferences. Navigate to the Voice tab. Toggle "Enable Voice Wake" to on.

Step 3: Choose your wake word. The default is "Hey Claw." You can select from presets (Hey Claw, OK Claw, Hello Claw) or enter a custom phrase. Custom phrases should be 2-4 syllables for reliable detection. Avoid words that sound similar to common speech to reduce false activations.

Step 4: Test it. Say your wake word clearly. You should hear a chime indicating Talk Mode is active. Speak a request — "What's on my calendar today?" — and wait for the response. Say "stop" or wait for the silence timeout to end the session.

Step 5: Fine-tune sensitivity. In the Voice preferences, adjust the wake word sensitivity slider. A value of 1 is very sensitive (triggers easily but may have false positives). A value of 10 is very strict (fewer false triggers but may miss your wake word in noisy environments). Start at 5 and adjust based on your experience.

Setting Up Voice on iOS

The OpenClaw iOS app brings voice capabilities to your iPhone and iPad. The setup is simpler than macOS since the app handles permissions during onboarding.

Step 1: Install the OpenClaw iOS app from the App Store. Connect it to your OpenClaw instance (either self-hosted or managed) by entering your server URL and authentication token.

Step 2: Enable Voice Wake. Open the app settings, go to Voice, and toggle "Voice Wake" on. Grant microphone permission when prompted.

Step 3: Configure and test. Choose your wake word, test it, and adjust sensitivity just like on macOS.

iOS limitation: Voice Wake only works when the OpenClaw app is in the foreground or recently backgrounded. iOS aggressively suspends background apps to preserve battery. For always-on voice, a Mac Mini is the better platform. On iOS, Talk Mode with manual activation (tap the microphone button) is the more practical approach for most users.

Talk Mode Deep Dive

Once Talk Mode is activated — whether by Voice Wake, tapping a button, or pressing a keyboard shortcut — a full voice conversation session begins.

How speech is processed:

1. Your speech is captured by the microphone.
2. Apple's Speech framework transcribes it to text on your device. No audio leaves your machine.
3. The text transcript is sent to OpenClaw, which processes it like any other message.
4. OpenClaw sends the text to your AI model provider (Claude, GPT, etc.) and receives a text response.
5. The response text is converted to speech using text-to-speech (Apple's native voices on macOS/iOS, Web Speech API in browsers).
6. The audio plays through your speakers or headphones.

Conversation continuity. Talk Mode maintains conversation context across multiple exchanges. You can have a multi-turn conversation: "What meetings do I have today?" followed by "Move the 2 PM meeting to 3 PM" — the agent understands "the 2 PM meeting" refers to the one it just told you about.

Keyboard shortcut (macOS). You can assign a global keyboard shortcut to toggle Talk Mode in the Voice preferences. Option+Space or Command+Shift+V are popular choices. This gives you push-to-talk without reaching for the mouse.

Silence timeout. After you stop speaking, Talk Mode waits for a configurable period (default 8 seconds) before ending the session. If you need more thinking time between exchanges, increase this to 15-20 seconds.

Advanced Configuration

Voice selection. macOS offers multiple text-to-speech voices. In Voice preferences, select your preferred voice. Premium voices (downloaded separately in macOS Accessibility settings) sound significantly more natural than the default voices. "Samantha (Enhanced)" and "Alex (Enhanced)" are popular choices for English.

Speaking rate. Adjust the speed of spoken responses from 0.5x (slow) to 2x (fast). Most users find 1.0x-1.2x comfortable for conversational interaction.

Audio routing. Choose which input device (microphone) and output device (speakers/headphones) Voice Wake and Talk Mode use. This is useful if you have multiple audio devices connected.

Do Not Disturb awareness. When enabled, Voice Wake automatically pauses during macOS Focus modes. This prevents activations during meetings, sleep hours, or other times when you do not want interruptions.

Voice response length. Configure whether the agent gives brief spoken responses (key points only) or detailed responses. For voice interactions, brief responses usually work better — you can always ask follow-up questions for details.

Real-World Use Cases

Morning briefing. Say "Hey Claw, give me my morning briefing" and hear a summary of today's calendar, important emails, pending tasks, and any overnight notifications. This takes 30 seconds versus 5-10 minutes of manual checking.

Hands-free task capture. While cooking, exercising, or driving, say "Hey Claw, remind me to call the dentist tomorrow at 10 AM." The agent adds it to your calendar or task list without you touching a device.

Meeting prep on the go. Walking to a meeting, say "Hey Claw, brief me on the Johnson account." The agent pulls up relevant notes, recent communications, and key talking points from your CRM and email history.

Email triage. "Hey Claw, summarize my unread emails." The agent reads through your inbox and gives you a voice summary of the important messages, who they are from, and what action is needed.

Quick calculations and research. "Hey Claw, what's 15% of $12,400?" or "Hey Claw, what time is it in Tokyo right now?" Quick questions answered instantly without switching apps or opening a browser.

Content dictation. "Hey Claw, draft a reply to Sarah's email about the project timeline. Tell her we'll have the deliverables ready by Friday and ask if she needs the specifications document before then." The agent drafts the email and can read it back for your approval.

Troubleshooting

Voice Wake not responding. Check that the OpenClaw app has microphone permission in System Settings. Verify Voice Wake is enabled in preferences. Test your microphone with another app. Reduce wake word sensitivity to a lower number (more sensitive). Speak clearly and at a normal volume — shouting or whispering can reduce detection accuracy.

False activations. Increase wake word sensitivity (higher number = stricter). Change your wake word to something more distinctive. Enable Do Not Disturb awareness to prevent activations during focused work.

Transcription errors. Apple's on-device speech recognition works best with clear speech and minimal background noise. Use an external USB microphone for better accuracy in noisy environments. Speak at a moderate pace — very fast speech increases transcription errors.

Slow responses. Voice latency comes primarily from the AI model API response time, not from speech processing. Switch to a faster model (Claude Haiku, GPT-4o-mini, Gemini Flash) for voice interactions where speed matters more than depth. Ensure your internet connection is stable — API calls require consistent connectivity.

Text-to-speech sounds robotic. Download enhanced voices in macOS System Settings under Accessibility, then Spoken Content, then System Voice, then Manage Voices. Enhanced voices are larger downloads (100-500MB) but sound significantly more natural.